Global warming can negate the expected CO2 stimulation in photosynthesis and productivity for soybean grown in the Midwestern United States.

Extensive evidence shows that increasing carbon dioxide concentration ([CO2]) stimulates, and increasing temperature decreases, both net photosynthetic carbon assimilation (A) and biomass production for C3 plants. However the [CO2]-induced stimulation in A is projected to increase further with warmer temperature. While the influence of increasing temperature and [CO2], independent of each other, on A and biomass production have been widely investigated, the interaction between these two major global changes has not been tested on field-grown crops. Here, the interactive effect of both elevated [CO2] (approximately 585 μmol mol(-1)) and temperature (+3.5°C) on soybean (Glycine max) A, biomass, and yield were tested over two growing seasons in the Temperature by Free-Air CO2 Enrichment experiment at the Soybean Free Air CO2 Enrichment facility. Measurements of A, stomatal conductance, and intercellular [CO2] were collected along with meteorological, water potential, and growth data. Elevated temperatures caused lower A, which was largely attributed to declines in stomatal conductance and intercellular [CO2] and led in turn to lower yields. Increasing both [CO2] and temperature stimulated A relative to elevated [CO2] alone on only two sampling days during 2009 and on no days in 2011. In 2011, the warmer of the two years, there were no observed increases in yield in the elevated temperature plots regardless of whether [CO2] was elevated. All treatments lowered the harvest index for soybean, although the effect of elevated [CO2] in 2011 was not statistically significant. These results provide a better understanding of the physiological responses of soybean to future climate change conditions and suggest that the potential is limited for elevated [CO2] to mitigate the influence of rising temperatures on photosynthesis, growth, and yields of C3 crops.

The global atmospheric concentration of CO 2 ([CO 2 ]) is predicted to rise from current concentrations of approximately 390 mmol mol 21 to between 730 and 1,020 mmol mol 21 by 2100 (Canadell et al., 2007;Intergovernmental Panel on Climate Change, 2007). An increase of [CO 2 ] is expected to stimulate photosynthesis, as has been demonstrated experimentally for a wide range of C 3 species (Curtis and Wang, 1998;Bernacchi et al., 2003a;Long et al., 2004;Nowak et al., 2004;Ainsworth and Long, 2005;Ainsworth and Rogers, 2007). Increasing [CO 2 ] coupled with the continued accumulation of other greenhouse gases in the atmosphere is predicted to increase the global average temperatures between 2.4°C and 6.4°C by the end of the century (Intergovernmental Panel on Climate Change, 2007). While the thermal optimum for photosynthesis ranges between 20°C and 35°C for C 3 species (Sage et al., 2008), plants have considerable capacity to acclimate to longterm temperature increases under current [CO 2 ]. Consequently, the combined changes in the atmospheric composition and climate are likely to impart significant effects on terrestrial ecosystems, because both CO 2 and temperature are critical determinants of photosynthetic rates (Sage and Kubien, 2007).
Soybean (Glycine max), grown in rotation with maize (Zea mays), represents the largest land use in the United States, combining to cover an estimated area of 67 million ha (National Agricultural Statistics Service, 2011). Soybean is the fourth most important commodity crop globally, with nearly 40% of world production coming from the midwestern United States. As one of the two crops that dominate the midwestern landscape, soybean strongly impacts regional ecosystem services, such as water quality, hydrologic cycling, and, as a grain legume, soil nitrogen production. Thus, global change-induced alterations in soybean production may have large-scale socioeconomic and ecological impacts.
Effective strategies to adapt agricultural production to global change require improved predictions of crop responses to global change scenarios. The Soybean Free Air CO 2 Enrichment (SoyFACE) research facility was developed to address the molecular, physiological, and growth responses of the midwestern crops soybean and maize to global change through the use of the Free Air CO 2 Enrichment technology (Miglietta et al., 2001). This technology provides in-field fumigation of CO 2 to simulate future atmospheric conditions while growing crops under otherwise natural field conditions using current agronomic practices. Previous results from SoyFACE showed that the daily integral of carbon uptake (A9) is increased by approximately 25% for soybean grown under elevated [CO 2 ] Bernacchi et al., 2006). This occurs despite the accompanying photosynthetic acclimation most evident as the reduction in the activity and/or content of the primary carboxylating enzyme Rubisco (Ainsworth and Long, 2005;Bernacchi et al., 2005). However, in soybean and other C 3 grain crops, CO 2 -induced increases in growth and yield are observed to be lower than the increase in net photosynthesis Long et al., 2006). In addition to increasing carbon assimilation (A) of soybean, elevated [CO 2 ] also reduces stomatal conductance (g s ) relative to plants grown in ambient [CO 2 ] (Ainsworth et al., 2002;Rogers et al., 2004;Bernacchi et al., 2006Bernacchi et al., , 2007. This reduction in g s can lead to lower canopy water use, improved water use efficiency, and conservation of soil moisture Bernacchi et al., 2007).
While Free Air CO 2 Enrichment experiments have improved our understanding of the plant physiological responses to a [CO 2 ]-enriched atmosphere, the interactive effects of the two major global change factors, rising [CO 2 ] and warming, have not been investigated by a direct manipulative experiment under field conditions for any crop. It has been shown that the stimulatory effect of elevated [CO 2 ] on A is enhanced at higher daily temperatures (Bernacchi et al., 2006), in agreement with theory (Long, 1991). However, this analysis was based on the natural variation in temperature during 15 d over three growing seasons, thus confounding the correlation with other environmental factors and neglecting long-term acclimation effects to higher temperature (Long, 1991;Sage and Kubien, 2007). Recent multiple regression analyses of historical yield data and growing season temperatures indicate that yields of soybean are depressed in warmer years (Lobell and Field, 2007;Kucharik and Serbin, 2008), with a steeper decline in yield above, compared with the incline below, a critical temperature (Schlenker and Roberts, 2009). It is also predicted that a 0.8°C increase in temperature could increase soybean yields in the midwestern United States by approximately 1.7% based on the mean air temperature of 22.5°C but could decrease yields for the warmer conditions in the southern United States (Hatfield et al., 2011). The results from this review (Hatfield et al., 2011) suggest that the impact of warming on the yield of soybean is highly dependent on baseline conditions. Because of the empirical nature of these analyses, the mechanisms involved in the decline of soybean productivity above the thermal optimum are unclear (Ainsworth and Ort, 2010). Accurately projecting the impact of global change on crop productivity relies on understanding how rising [CO 2 ] and increasing temperature together influence the photosynthesis, growth, phenology, and yield of major crop species.
The objectives of this study were to quantify and understand the biological processes involved in photosynthesis, growth, and biomass production for soybean under a warmer and CO 2 -enriched atmosphere. Specifically, we hypothesized that increasing the growing season temperature of a soybean canopy by 3.5°C above the current ambient conditions, an increase expected midway through this century for terrestrial areas (Rowlands et al., 2012), will result in lower photosynthesis, biomass productivity, and yields. This is consistent with what has been observed historically for warmer years. We further hypothesized that the combination of elevated [CO 2 ] and warmer temperatures will yield higher photosynthesis, biomass accumulation, and yield relative to the elevated [CO 2 ] treatment alone. Given the high interannual variability in climate, these hypotheses were tested over two growing seasons; we hypothesized that the responses to elevated [CO 2 ], warmer temperature, and the combined treatment would be consistent across growing seasons.

Meteorological Conditions Contrasted between the 2009 and 2011 Growing Seasons
The mean temperatures for July, August, and September were lower in 2009 and higher in 2011 when compared with the 30-year mean ( Table I). The largest departure from normal in 2009 was observed in July, with temperatures approximately 2.9°C below average. In 2011, July had the highest departure, with mean temperatures 2.8°C above average. In both years, April and May experienced higher than average precipitation, ensuring that both growing seasons began with soil moisture near field capacity. The entire 2009 growing season experienced near-average precipitation, whereas in 2011, there were significant shortfalls of 60% in July and 47% in August compared with the 30-year means for those months (Table I). Growth stages (Supplemental Table S1) and specific meteorological conditions ( Fig. 1 Measurements of A were collected throughout the daylight hours on seven days (2009) and six days (2011) throughout each growing season. The measurements were taken in conditions set to mimic the field conditions at each specific time (Supplemental Figs. S1 and S2). Diurnal measurements showed variable responses of the key gas-exchange parameters A, g s , intercellular [CO 2 ] (C i ), and intrinsic water use efficiency (iWUE) over the course of the two growing seasons (Supplemental Tables S3 and S4). Relative to the control, the elevated [CO 2 ] and ambient temperature (eC) plots yielded higher A, the elevated temperature and ambient [CO 2 ] (eT) plots yielded lower A, and the elevated temperature and elevated [CO 2 ] (eT+eC) plots yielded higher A over most of the time points (Supplemental Figs. S1 and S2).
The diurnal measurements of A were integrated across the daylight hours to obtain A9. Overall, the effect on A9 of eT and of eC relative to the control plots was consistent across both growing seasons, although the increase in eC and the decrease in eT were greater in 2011 (Fig. 2). In 2009, there was a statistically  significant temperature 3 CO 2 interaction. This interaction is explained by a lack of statistical difference between the control and eT plots, while the eT+eC plot had much higher A9 relative to the control (Tables II  and III). In 2009, both eC and eT+eC treatments had significantly higher A9 than the ambient [CO 2 ] treatments (i.e. control and eT; Fig. 2; Table II), although eC and eT+eC were not detectably different from each other (Table III). The responses of A9 among the treatments in 2009 did not result in statistically significant interactions of main effects with day of the year (DOY; Table II; Supplemental Table S3).
In contrast to 2009, in 2011, there were no statistically significant interactive effects of temperature and [CO 2 ] on A9, suggesting that the main effects alone were responsible for the observed differences. Elevated [CO 2 ] increased A9 within a temperature treatment (eC versus control and eT+eC versus eT), and elevated temperature decreased A9 within a CO 2 treatment (eT versus control and eT+eC versus eC; Tables II and III; Fig. 2). The CO 2 effect, based on percentage change, was greater than the temperature effect; thus, the combined influence of increasing temperature and rising [CO 2 ] stimulated A9 in the eT+eC treatment relative to the control, but the increase was lower than the eC versus control comparison (Table III; Fig. 2). This response was different from the pattern observed in 2009, where A9 in eT+eC showed the highest value, although significantly different from eC. The effect of temperature varied by DOY in 2011, when temperature did not have significant effects on A9 on two measurement dates (DOY 186 and 251; Fig. 2; Supplemental Table S4).
Over both years, elevated [CO 2 ] and warming reduced g s when treatments were applied independently and in combination ( Fig. 2; Table II), with the eT+eC treatment showing the lowest g s overall for both years (Table III; Fig. 2). The influence of CO 2 relative to temperature, however, appeared to change from 2009 to 2011. In 2009, elevated [CO 2 ] had a greater impact on g s compared with temperature, whereas in 2011, temperature had a bigger impact on g s than [CO 2 ] (Fig.  2). The impact of elevated temperature and/or [CO 2 ] on g s varied based on DOY for both years (Table II;  Supplemental Tables S3 and S4). The [CO 2 ] treatment differences were statistically significant on all but one day for each of the years. The temperature-induced reduction in g s was statistically significant on four of  Table III; Supplemental  Tables S3 and S4).
Over both growing seasons, elevated temperatures reduced and elevated [CO 2 ] increased C i relative to the controls (Tables II and III; Fig. 2). While the pattern of responses was similar over both growing seasons, the influence of temperature in 2011 was much greater than that observed in 2009. Despite a significant temperature 3 CO 2 interaction in 2011, the increase in C i for the eC treatment and the decrease in C i for the eT treatment appeared to have an additive effect on the eT+eC treatment for both years.
The percentage change in A9 within a [CO 2 ] level (eT versus control and eT+eC versus eC) was plotted as a function of the daily maximum canopy temperature based on the temperatures measured in the heated plots (Fig. 3). The percentage change in A9 between the eT+eC and eC treatments increased with temperature in 2009 and decreased with temperature in 2011 (Fig.  3). The decline in A9 for eT+eC relative to eC in 2011 was rapid and stabilized at daily maximum temperatures above 34°C. The eT treatment decreased A9 slightly relative to the control in 2009 and remained about 20% lower for all daily temperatures in 2011 (Fig. 3).
The relationship of instantaneous A to g s increased to an asymptote for all treatments (Fig. 4). While it was difficult to discern differences within a [CO 2 ] treatment (eT versus control and eT+eC versus eC), it was clear that for a given g s , the elevated [CO 2 ] treatments had much higher A. Averaging A and g s within each measurement day for both years to calculate the daily mean and seasonal mean iWUE (A/g s ) indicated significant differences among the treatments (Table II). For both years, there were statistically significant increases in iWUE associated with elevated temperature and with [CO 2 ] treatment (Table III; Fig. 4). While the responses of iWUE were similar for both growing seasons, the CO 2 effect was greater in 2009 than in 2011 and the temperature effect was greater in 2011 than in 2009 (Table III; Fig. 4). For both years, it appears that the influence of the two main effects was additive (Table III; Fig. 4).

Elevated Temperature Slightly Reduced Leaf Water Potential and Increased Leaf-to-Air Vapor Pressure Deficit
Leaf water potential estimated using thermocouple psychrometry was determined on leaf tissue collected at midday of each measurement date. Temperature, regardless of the [CO 2 ], had a small but significant effect on total water potential (WP) and turgor pressure (TP) in both years ( Fig. 5; Table II). Elevated temperature decreased growing season mean WP by 7% in 2009 and by 16% in 2011 (Table III). While no statistically significant results were observed with osmotic potential (OP) for either year, the temperature main effect decreased TP by 17% in 2009 and 16% in 2011. The seasonal and daily mean values of WP, OP, and TP in 2011 were nearly one-half the values obtained in 2009 (Fig. 5).
The seasonal mean leaf-to-air vapor pressure deficit (VPD) obtained from the gas-exchange system during the in situ measurements showed higher values in the warmer (1.58 6 0.022 kPa for eT, 1.77 6 0.023 kPa for eT+eC) relative to the reference (1.27 6 0.018 kPa for control, 1.49 6 0.020 kPa for eC) temperature plots in 2009. The relative responses in 2011 were similar to those in 2009, with the heated plots showing higher VPD (1.93 6 0.030 kPa for eT, 2.09 6 0.033 kPa for eT+eC) relative to the reference temperature plots (1.51 6 0.025 kPa for control, 1.68 6 0.026 kPa for eC). These values gave percentage deviations from control of 24.7% for eT, 17.6% for eC, and 39.6% for eT+eC in 2009 and 28.1% for eT, 11.4% for eC, and 38.4% for eT+eC in 2011.
Elevated CO 2 Increased Biomass and Yields, But the Impact of Temperature Varied with Growing Season Both growing seasons resulted in statistically significant CO 2 3 temperature interactions for aboveground biomass (AGB), indicating that a synergistic effect occurred between these two global changes. The treatment responses, however, were not consistent over both growing seasons. In 2009, the only two differences among the treatments were between the eT+eC comparison with the control plot and the eT+eC comparison with the eT plot (Tables II and III; Fig. 6). In 2011, AGB in the eC plots were higher than the control and eT+eC plots (Tables II and III; Fig. 6). The general responses of seed yield (SY) for both growing seasons were relatively similar to the observations of AGB; however, in 2009, the only statistically significant differences occurred between the eT+eC and eT treatments, with the eT treatment showing the lowest SY and the eT+eC treatment showing the highest SY (Tables II and III ; Fig. 6). The impact of temperature was strongly negative on yields in 2011, with the eT and eT+eC plots having lower SY compared with both the control and eC plots (Tables II and III; Fig. 6).
All treatments appeared to decrease the harvest index (HI) in both years (Fig. 6); however, in 2011, the only statistically significant differences were between eT+eC and the control (Table II). Overall, the HI in 2011 was much lower relative to 2009, and the response of HI to the various treatments was amplified. In 2011, neither eT nor eC differed from the control plot, but the eT+eC treatment differed from the control, eT, and eC plots (Table II). The differences between the treatments and the control in 2011 were greater than in 2009, although there was also much greater variance in the measurements (Fig. 6).

DISCUSSION
This study was designed to test the hypotheses that (1) increasing soybean canopy temperature above ambient will result in lower photosynthesis, biomass productivity, and yields, an effect that is historically characteristic of warmer years; and (2) concomitant increases in [CO 2 ] and temperature will result in higher photosynthesis, biomass production, and yields compared with elevated [CO 2 ] alone. The results only partially support the first hypothesis. Significant differences between the control and eT treatments were observed only for photosynthesis (Fig. 2) and SY (Fig.  6) in 2011, the warmer of the two growing seasons. The results also indicate that the second hypothesis is not supported. In 2009, the eT+eC treatment yielded significantly greater A on only two sampling days relative to the eC treatment, and in 2011, the eT+eC treatment yielded significantly lower photosynthesis than the eC treatment. Moreover, in 2011, the response of photosynthesis for the eT+eC treatment, which was slightly higher than for the control plots, did not translate to increases in SY; rather, the eT+eC plot had lower SY than any other treatment. The combined eT+eC treatment yielded higher photosynthetic rates in both years compared with the eT treatment, suggesting that eC is able to at least partially mitigate the negative effects of eT, but this response did not lead to higher SY for the eT+eC plot compared with the eT plot in 2011. In 2011, the eT plots, with (eT+eC) or without (eT) elevated [CO 2 ], yielded lower SY relative to the control.
The influence of rising [CO 2 ] on photosynthesis, growth, and yield has been well documented for soybean Bernacchi et al., 2006) as well as for other species (Bernacchi et al., 2003a;Long et al., 2004;Nowak et al., 2004;Ainsworth and Long, 2005;Ainsworth and Rogers, 2007). Consistent with previous research on soybean grown at SoyFACE (Leakey et al., 2009), significant increases in photosynthesis in elevated [CO 2 ] were observed in both years (Tables II  and III). The 2009 growing season was the only time out of 10 growing seasons at SoyFACE in which increases in SY with elevated [CO 2 ] were absent. One of the key benefits of an increase in photosynthesis under elevated [CO 2 ] is the reduction of photorespiration (Long, 1991;Long et al., 2004;Ainsworth and Rogers, 2007). The smaller than predicted response in 2009 could be attributed to cooler temperatures, where the inhibition of photosynthesis due to photorespiration was low (Jiao and Grodzinski, 1996). This, coupled with a large decrease in g s (Fig. 2) under elevated [CO 2 ], could have contributed to the muted yield response in 2009. The 2011 growing season was warmer than typical; thus, the ability to suppress photorespiration at high [CO 2 ] likely conferred a greater benefit under these conditions. The 2011 growing season showed higher biomass and yields for the elevated-[CO 2 ]-grown plants, as observed for most growing seasons at SoyFACE.
The heated plots showed consistently reduced photosynthetic carbon uptake for both growing seasons relative to the control. It was clear that the increase in temperature had a greater effect on reducing photosynthesis in 2011 than in 2009, likely driven by the much warmer background temperatures in 2011. This is observed in the seasonal percentage deviation of A9 for eT versus control, which showed a slight decrease in 2009 and a much larger decrease in 2011 (Table III). This was also seen in the percentage deviation of A9 versus daily maximum temperature for eT versus control (Fig. 3). In 2009, as temperature in the heated plots increased, the percentage difference of A9 between the eT+eC and eC plots continued to increase. Contrary to this, the differences between the eT versus control plots declined, indicating greater heat-induced suppression of A9 with temperature. In 2011, A9 was consistently lower in the eT treatment relative to the control by approximately 20% regardless of daily maximum temperature, whereas the stimulation associated with the eT+eC plot dropped as temperatures rose (Fig. 3). These results suggest that in 2009, photosynthesis in the nonheated plots was below the thermal optimum, driving photosynthesis higher as heat was applied, whereas in 2011, photosynthesis for the nonheated plots was operating above the thermal optimum, such that any further increases in temperature drove photosynthesis down.
The potential importance of the interaction of elevated [CO 2 ] and increased temperature on photosynthesis is well described using mechanistic theory (Long, 1991;Sage and Kubien, 2007) and demonstrated using correlative data collected over many days over different growing seasons (Bernacchi et al., 2006). The results presented here show that under field conditions, the importance of the elevated [CO 2 ] and temperature effect on photosynthesis, growth, and yield is dependent on the extent of heating, on whether the temperature was above or below optimum at any particular time, and on the interaction with other environmental factors. Of particular importance is the role of changes associated with the underlying biochemistry of photosynthesis. Elevated [CO 2 ] is shown to down-regulate the maximum velocity for carboxylation (V c,max ) for soybean grown in elevated [CO 2 ]; however, this acclimation had little impact on photosynthetic rates, as soybean is not V c,max limited in elevated [CO 2 ] (Bernacchi et al., 2005). Acclimation has been shown to occur in response to higher growth temperature for V c,max and maximum rate of electron transport (J max ; June et al., 2004;Onoda et al., 2005;Kattge and Knorr, 2007). Over the same growing seasons and coupled with the measurements presented here, in-depth analysis of V c,max and J max for soybean at SoyFACE To determine whether photosynthesis was predominately Rubisco or ribulose-1,5-bisphosphate (RuBP) regeneration limited over each diurnal for all treatments, we coupled the results from D.M. Rosenthal, U.M. Ruiz-Vera, M. Siebers, C.J. Bernacchi, and D.R. Ort (unpublished data) for V c,max and J max with the leaf photosynthesis model (Farquhar et al., 1980) corrected for temperature (Bernacchi et al., 2001(Bernacchi et al., , 2003b. Using this modeled data, we determined whether photosynthesis is Rubisco or RuBP regeneration limited for all data points collected over the two growing seasons. This analysis showed that over 85% of all data points were RuBP limited (Supplemental Figs. S1 and S2). Because Rubisco-limited photosynthesis was rare, any down-regulation in V c,max is likely to not influence A. Given its importance for RuBP regeneration-limited A, the down-regulation of J max is likely to drive down productivity.
The responses of photosynthesis to elevated [CO 2 ] and warmer temperature over these two growing seasons are not likely driven exclusively by the temperature and CO 2 sensitivity of Rubisco kinetics and RuBP regeneration (D.M. Rosenthal, U.M. Ruiz-Vera, M. Siebers, C.J. Bernacchi, and D.R. Ort, unpublished data). Previous research at SoyFACE showed that elevated [CO 2 ] resulted in an approximately 16% mean reduction in g s (Bernacchi et al., 2006), which is substantially less than the reduction observed here (Table  III). Thus, the pronounced reduction of g s under elevated [CO 2 ] and/or temperature over both years is likely to influence the deviation of the observations from theoretical responses. Over two sampling dates in 2009 (DOY 183 and 238), the differences in g s were minimal for the eT+eC versus eC treatments (Fig. 2). The combined increase in [CO 2 ] and temperature for these two days yielded much higher A9, consistent with the theory presented previously (Long, 1991). All other measurement days during 2009 yielded g s values that were lower in the eT+eC relative to the eC treatment and did not show differences in A9 between these treatments. The temperature-induced reduction in g s led to lower C i within a CO 2 treatment (Fig. 2). The effect of temperature resulted in a much greater decrease for C i in 2011 than in 2009. In 2011, C i in the eT+eC treatment was only approximately 20% higher than the control despite atmospheric CO 2 being approximately 50% higher.
Despite the reductions in g s , the eT, eC, and eT+eC treatments showed higher iWUE ( Fig. 4; Table III). This increase in iWUE is predicted for plants grown in elevated [CO 2 ], as stomatal limitation to photosynthesis is consistently shown to be lower despite a decrease in g s (Bernacchi et al., 2005;D.M. Rosenthal, U.M. Ruiz-Vera, M. Siebers, C.J. Bernacchi, and D.R. Ort, unpublished data). The elevated-temperature-grown plants also showed an increase in iWUE across both growing seasons relative to the control. The increase in iWUE for the heated treatments within a [CO 2 ] (eT versus control and eT+eC versus eC) was less than that measured within a temperature treatment (eC versus control and eT+eC versus eC). The increase associated with the eT treatment relative to the control occurred as a result of the decrease in g s being proportionately greater than the decrease in A9 (Table III). This indicates that while all treatments experienced a higher iWUE, the mechanisms behind these responses varied among treatments.
A characteristic of an experiment that heats the canopy instead of the air is that it will lead to an increase in VPD at the leaf surface, which likely leads to an increase in water use (Kimball, 2005(Kimball, , 2011De Boeck et al., 2012). Averaged across time points in which gasexchange measurements were collected, the effect of heating increased the leaf-to-air VPD by 0.3 kPa in 2009 and 0.4 kPa in 2011 within a CO 2 treatment (eT versus control and eT+eC versus eC). Both growing seasons experienced significant precipitation during spring, and 2009 had ample precipitation throughout the season (Table I). Leaf water potential data for both years showed the same responses to temperature, with a general response of temperature within a CO 2 treatment showing more negative WP relative to the nonheated reference plots (Fig. 5; Table II). The differences in WP, OP, and TP in 2011 compared with 2009 indicate that the variability in climate between the two growing seasons had a dominant influence on leaf water potential relative to the treatment imposed by the infrared heating arrays. Moreover, pairwise comparison analysis showed no statistical differences among treatments for the WP and TP variables in either year, suggesting that the photosynthetic and productivity values obtained for 2011 are representative for the environmental conditions tested and not artificially altered by the experiment. Thus, it is not likely that the additional water use in the heated plots had a substantial effect on growth and physiology during 2009.
With the exception of the eT+eC treatment relative to the control in 2011, the responses observed for A9 were similar to those observed for AGB. It is interesting, however, that the response of SY to the various treatments did not follow the responses of A9 and of AGB. In 2009, the AGB was significantly higher in the eT+eC treatment relative to the control, but the SY for this treatment did not differ statistically from the control. The statistically significant increase in SY for the eT+eC relative to the eT treatments was the only statistically significant difference in SY in 2009. This suggests that, in this year, the increase in [CO 2 ] offset the losses typically associated with an increase in temperature. In 2011, there were no differences in SY between the eT+eC and eT treatments, which implies that the benefit of both factors together is not universal. In 2011, increases occurred for both AGB and SY in the eC treatment relative to the control, but the loss in SY was amplified in relation to the loss in AGB for both heated treatments (Table III).
Within a growing season, increasing temperature appears to reduce yields (Table III; Fig. 6), despite the 13.5% decrease in 2009 not being statistically different from the control (P , 0.15). Recent multiple regression analyses of historical yield data and growing season temperatures indicate a negative relationship, meaning that yields of soybean are depressed in warmer years (Lobell and Field, 2007;Kucharik and Serbin, 2008), a phenomenon that is consistent with our findings. A separate analysis, however, suggests that increases in yields with rising temperatures are likely to occur in cooler areas (e.g. midwestern United States) while decreases in yields with temperature are to occur in traditionally warmer areas (e.g. southern United States; Hatfield et al., 2011). Our data indicate that additional heating did not confer an advantage to SY, even in a year when background temperatures were much cooler than the long-term mean (2009). A problem with historical trends is that they neglect other confounding factors that cannot be controlled for, such as an interaction with an increase in [CO 2 ]. When the interaction between rising [CO 2 ] and temperature was considered (Fig. 6), it was clear that the yield responses were not consistent with the observed photosynthetic responses and that they varied based on growing season conditions. The results from the 2009 growing season suggest that the increased [CO 2 ] completely negated the detriment of increased temperature on yield. The 2011 growing season showed that the addition of [CO 2 ] with higher temperatures did nothing to mitigate the influence of higher temperature. As the conditions associated with the 2011 growing season are predicted to become more common (Hayhoe et al., 2010), these results suggest that the assumptions, based on the theory that rising [CO 2 ] and increasing temperature could synergistically increase yields, need to be reassessed.
All treatments in 2009 and the elevated-temperature treatment in 2011 resulted in decreases in HI compared with the control. The decline in HI for elevated-[CO 2 ]grown soybean is consistent with a number of previous studies on soybean (Amthor et al., 1994;Heagle et al., 1998;Ziska and Bunce, 2000;Morgan et al., 2005). Similarly, the influence of increasing temperatures drives the HI for soybean down, with a decrease in seed size being a major factor for this response (Boote et al., 2005). All treatments in 2009 had a smaller percentage decrease on HI compared with the control in 2009 than in 2011 (Fig. 6). This is consistent with previous reports for soybean that show an accelerated decline in HI as temperatures increase (Boote et al., 2005). While higher temperatures were not directly imposed upon the eC treatments in 2011, the CO 2 -induced closure of g s warmed canopies above the temperatures in the control plots, as reported previously (Bernacchi et al., 2007). The measured canopy temperature data used in the heating control system (Supplemental Fig. S4) indicate that season-mean increases in CO 2 resulted in approximately 1°C warmer canopy temperatures during midday hours. This CO 2induced warming, which was apparent in 2011 but not in 2009, could potentially contribute to the larger decline in HI for the elevated-[CO 2 ] treatment relative to the control. There were marked differences in biomass components, leaf physiology, and water potential between the 2009 and 2011 growing seasons. The factors that can influence each of these processes are complex and subject to a significant number of environmental factors. A full understanding of the drivers behind the differences between these two growing seasons would need to account for these factors and require analyses that extend beyond the measurements in this paper.

CONCLUSION
The results from this research show that increased temperatures reduced photosynthesis, growth, and yield in soybean. These responses were linked with reductions in g s , although biochemical changes to photosynthesis (D.M. Rosenthal, U.M. Ruiz-Vera, M. Siebers, C.J. Bernacchi, and D.R. Ort, unpublished data) likely also influence these photosynthetic results. Moreover, the combined effects of elevated [CO 2 ] and warmer temperature did not lead to significant increases in photosynthesis compared with an increase in [CO 2 ] alone. In fact, the combined increases in [CO 2 ] and temperature led to reduced photosynthesis relative to elevated [CO 2 ] alone in one of the two years. The contrasting climatic conditions over the two years of measurements played a significant role in the different photosynthetic, growth, and yield rates observed. These results suggest that in future climate change conditions, the interactive effects of elevated [CO 2 ] and warmer temperatures will likely not benefit soybean physiology, growth, and development as predicted from theory, due to overriding environmental factors.

Site Description and Experimental Design
This experiment was conducted on soybean (Glycine max 'Pioneer 93B15') during the 2009 and 2011 growing seasons as part of the Temperature 3 Free Air CO 2 Enrichment experiment located at the SoyFACE research facility in Champaign, Illinois. The entire SoyFACE research farm consists of a 32-ha (80acre) field in Illinois (40°2'30.49'' N, 88°13'58.80'' W, 230 m above sea level). Characteristics of the site and details of agriculture practices can be found in papers by Ainsworth et al. (2004), Rogers et al. (2004), andBernacchi et al. (2006). The experiment consisted of a randomized complete block design with four blocks to account for field topographic and soil variations. Each block contained one control and one elevated-[CO 2 ] plot; each plot had a diameter of 20 m, and they were separated from each other by 100 m as described previously (Miglietta et al., 2001). Each heated plot contained six infrared heaters (Salamander Aluminum Extrusion Reflector Assembly Housing for Ceramic Infrared Heaters; Mor Electric Heating) each fitted with four infrared heating elements (Mor-FTE 1,000-W, 240-V heaters; Mor Electric Heating Association). The six heaters were arranged in a 3-m-diameter hexagonal pattern with a total heating area of 7.1 m 2 (Supplemental Fig. S3). The heaters were maintained 1.2 m above the canopy, tilted toward the center of the plot at a 45°angle as described by Kimball et al. (2008). The heater output was regulated using a custom-built industrial dimmer system in which two thyristors (dimmers) were controlled using one circuit board, all of which were taken from one complete dimmer assembly (model LCED-2484, 240 V, 35 A;Kalglo Electronics). The dimmer circuit board controlled a range of output up to 24,000 W of infrared heating power. The actual 0-to 240-V alternating current output was scaled from a 0-to 10-V direct current input signal to the dimmer. Each heated plot was controlled using a datalogger (CR1000 Micrologger; Campbell Scientific). The datalogger used a proportional-integrative-derivative feedback control system, similar to the ones used by Kimball (2005), to maintain an approximately 3.5°C increase over the ambient temperature 24 h per day throughout the growing seasons (Supplemental Fig. S4). The reference and heated canopy temperatures were measured using infrared radiometers (SI-121; Apogee Instruments) wired into the datalogger equipped with a voltage-output module (SDM-CV04; Campbell Scientific). Based on the canopy temperature difference between the heated versus reference plots, the voltage output module supplied the 0-to 10-V signal to the dimmer to maintain the target temperature rise of heated plots above the reference plots. The full experiment consisted of four treatments: control (ambient [CO 2 ] and ambient temperature), eT (ambient [CO 2 ] and +3.5°C in temperature), eC (585 mmol mol 21 [CO 2 ] and ambient temperature), and eT+eC (585 mmol mol 21 [CO 2 ] and +3.5°C in temperature).

Meteorological and Micrometeorological Data
Temperature, humidity, and solar radiation for the research site were obtained from a meteorological station associated with the Surface Radiation Network (40.05N, 88.37W; http://www.srrb.noaa.gov/surfrad/index.html) and processed as described by Vanloocke et al. (2010). Precipitation was obtained from the Willard airport station (40.04N, 88.27W; http://cdo.ncdc. noaa.gov/qclcd/QCLCD). A record for 30-year mean temperature and precipitation for zone 5 of the Illinois climate division (which corresponds to Champaign) was obtained from the Midwestern Regional Climate Center (http://mrcc.isws.illinois.edu/).

Gas-Exchange Measurements and Midday Sampling
Diurnal measurements of instantaneous A and other physiological data, such as C i and g s , were collected using portable open gas-exchange systems with incorporated infrared CO 2 and water vapor analyzer (Li-Cor 6400; Li-Cor) coupled with an integrated chlorophyll fluorometer (LI-6400-40 leaf chamber fluorometer; Li-Cor). The infrared CO 2 and water vapor analyzers were zeroed, and the gas-exchange systems were calibrated as described by Bernacchi et al. (2006).
Gas-exchange sampling occurred every 2 weeks throughout both growing seasons starting with the V3 vegetative developmental stage (i.e. following the emergence of the third trifoliate) and ending with the approximately R6 developmental stage (full seed) based on the development classifications of Ritchie et al. (1993). There were a total of seven measurement days in 2009 and six in 2011 (Supplemental Table S1). The measurements were taken at 2-h intervals between 9 AM and 5 PM on three plants per plot from the youngest fully expanded leaves. Four research teams each using matched instrumentation conducted measurements at each time point of the diurnal in different blocks to sample all 16 plots within 45 min. The order of sampling for each team was randomized among blocks. At the beginning of each time point measurement, photosynthetic photon flux density (LI-190; LI-COR) and air temperature (HMP-45C; Campbell Scientific [mounted in aspirated temperature shield model 076B; Met One Instruments]) were recorded from sensors located at SoyFACE. The block temperature of the gas-exchange systems was set according to ambient air temperature and adjusted to higher temperature (ambient + 3.5°C) for the heated plots. The sample relative humidity was between approximately 50% and 70%. The [CO 2 ] in the gas-exchange system reference chamber was set to concentrations that corresponded to the control (400 mmol mol 21 ) or elevated (600 mmol mol 21 ) [CO 2 ] plots. These concentrations, while above the mean concentrations in each plot, allowed for the leaf to draw CO 2 down in the sample chambers to closely match plot means and accounted for the difference in the control and elevated [CO 2 ] for 2009 compared with 2011. The values of A, g s , and C i were calculated using the integrated software in the gas-exchange system according to von Caemmerer and Farquhar (1981). The A9 was calculated from instantaneous A as described by Leakey et al. (2004).
Values for V c,max , J max , and respiration in the light at 25°C were obtained from A versus C i curves measured within 1 to 2 d of each diurnal for both growing seasons (D.M. Rosenthal, U.M. Ruiz-Vera, M. Siebers, C.J. Bernacchi, and D.R. Ort, unpublished data). These parameters were used to model A using the leaf photosynthesis model (Farquhar et al., 1980) to determine which process, Rubisco or RuBP regeneration, was limiting photosynthesis at each time point for all treatments. The model was corrected for measured leaf temperature using the temperature functions provided previously (Bernacchi et al., 2001(Bernacchi et al., , 2003b. Percentage stimulation of A9 per plot for each day was calculated with the plot mean values for eT versus control and for eT+eC versus eC. The percentage stimulation was plotted as a function of the daily maximum temperatures for the heated plot. The values used were from four measurement days in 2009 (DOY 197, 210, 224, and 238) and 2011 (DOY 200, 214, 228, and 242), selected according to similarities in the VPD and solar radiation and adequate canopy cover to prevent the influence of soil in the measurements of canopy temperature (Leaf Area Index . 2). The iWUE was calculated with A O g s .
Samples to determine leaf WP were collected during the midday time point on each measurement day with five subsamples taken per plot. Three leaf tissue discs of 1.2-cm diameter were excised per plant and sealed in psychrometer chambers (C-30; Wescor). The chambers were equilibrated in a controlled-environment growth chamber at 25°C as described previously . After thermal equilibration, WP was measured using a dew-point microvoltmeter (HR-33T; Wescor) integrated into the psychrometers. Upon measuring WP, the chambers were submerged into liquid nitrogen, and the potentials were recorded on the lysed plant tissue to determine OP. The TP was calculated as WP -OP. A calibration standard was obtained independently each year using Suc solutions ranging in concentration from 0 to 1.60 M.

AGB and Yield Measurements
AGB was obtained at the end of each growing season after full maturity (growth stage R8) was reached (DOY 267 in 2009 andDOY 298 in 2011). Five 1.5-m rows of soybeans per plot in 2009 and two 1.0-m rows per plot in 2011 were harvested by hand. At harvest, plants had few attached leaves, so AGB included only pods and stems. These biomass components are typically used for determining HI. Each component was dried to constant weight (approximately 7 d) at 65°C and weighed. The pod tissues were then run through a thresher to isolate the seeds, and the seeds were weighed to obtain SY. The HI was calculated by dividing SY by AGB.

Statistical Analysis
Photosynthesis, water potential, and iWUE were analyzed using a complete block repeated-measures mixed-model ANOVA using the PROC MIXED command with the Kenward-Roger method in SAS System 9.3 (SAS Institute). The fixed effects for the seasonal analysis were as follows: DOY, [CO 2 ], temperature, temperature 3 [CO 2 ] interaction, and DOY with the interaction of the other fixed effects. Statistical tests of within-day differences among the treatments were analyzed separately for each day using time of day, rather than DOY, as the repeated factor in the analysis. The AGB, yield data, and HI were analyzed similarly to the previous variables but without including DOY in the analysis. The differences of least-square means from Student's t tests were used to compare individual treatment means; this option is integrated into SAS System 9.3. To lower the possibility of a type II error, the statistical significance was evaluated at a # 0.1.

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S1. Diurnal photosynthesis data for each measurement day in 2009. Figure S2. Diurnal photosynthesis data for each measurement day in 2011.

Supplemental
Supplemental Figure S3. Aerial view of experiment infrastructure.
Supplemental Figure S4. Temperature 3 Free Air CO 2 Enrichment temperature control data for 2009 and 2011.
Supplemental Table S1. Timeline of planting, measurement dates, and final harvest with corresponding development stages.
Supplemental Table S2. Meteorological data for each measurement date in 2009 and 2011.
Supplemental Table S3. Per measurement day statistical tables for each variable measured in 2011.